Test

In your final repo, there should be an R markdown file that organizes all computational steps for evaluating your proposed Facial Expression Recognition framework.

This file is currently a template for running evaluation experiments. You should update it according to your codes but following precisely the same structure.

if(!require("EBImage")){
  install.packages("BiocManager")
  BiocManager::install("EBImage")
}
if(!require("R.matlab")){
  install.packages("R.matlab")
}
if(!require("readxl")){
  install.packages("readxl")
}

if(!require("dplyr")){
  install.packages("dplyr")
}
if(!require("readxl")){
  install.packages("readxl")
}

if(!require("ggplot2")){
  install.packages("ggplot2")
}

if(!require("caret")){
  install.packages("caret")
}

if(!require("glmnet")){
  install.packages("glmnet")
}

if(!require("WeightedROC")){
  install.packages("WeightedROC")
}

if(!require("e1071")){
  install.packages("e1071")
}

if(!require("xgboost")){
  install.packages("xgboost")
}

if(!require("randomForest")){
  install.packages("randomForest")
}
if(!require("gbm")){
  install.packages("gbm")
}

library(R.matlab)
library(readxl)
library(dplyr)
library(EBImage)
library(ggplot2)
library(caret)
library(glmnet)
library(WeightedROC)
library(e1071)
library(xgboost)
library(randomForest)
library(gbm)

Step 0 set work directories

set.seed(2020)
# setwd("~/Project3-FacialEmotionRecognition/doc")
# here replace it with your own path or manually set it in RStudio to where this rmd file is located. 
# use relative path for reproducibility

Provide directories for training images. Training images and Training fiducial points will be in different subfolders.

test_dir <- "../data/test_set_predict/" # This will be modified for different data sets.
test_image_dir <- paste(test_dir, "images/", sep="")
test_pt_dir <- paste(test_dir,  "points/", sep="")
test_label_path <- paste(test_dir, "label_prediction.csv", sep="") 

Step 1: set up controls for evaluation experiments.

In this chunk, we have a set of controls for the evaluation experiments.

• (T/F) cross-validation on the training set
• (T/F) reweighting the samples for training set
• (number) K, the number of CV folds
• (T/F) process features for training set
• (T/F) run evaluation on an independent test set
• (T/F) process features for test set

run.cv.baseline <- FALSE # run cross-validation on the gbm baseline
sample.reweight <- TRUE # run sample reweighting in model training
K <- 5  # number of CV folds
run.feature.train <- TRUE # process features for training set
run.test <- TRUE # run evaluation on an independent test set
run.feature.test <- TRUE # process features for test set
run.cv.xgboost <- FALSE

Using cross-validation or independent test set evaluation, we compare the performance of models with different specifications. In this Starter Code, we tune parameter lambda (the amount of shrinkage) for logistic regression with LASSO penalty.

#gbm parameters tuning:
n.trees = c(10,50,100,200)
shrinkage = c(0.01,0.05,0.1,0.15)

#xgboost parameters tuning
n_iterations <- 100

Step 2: import data and train-test split

set.seed(2020)
#train-test split
info <- read.csv(test_label_path)
n <- nrow(info)
#train_idx <- sample(info$Index, n_train, replace = F)
test_idx <- seq(1, n, by=1)

If you choose to extract features from images, such as using Gabor filter, R memory will exhaust all images are read together. The solution is to repeat reading a smaller batch(e.g 100) and process them.

n_files <- length(list.files(test_image_dir))

image_list <- list()
for(i in 1:100){
   image_list[[i]] <- readImage(paste0(test_image_dir, sprintf("%04d", i), ".jpg"))
}

Fiducial points are stored in matlab format. In this step, we read them and store them in a list.

#function to read fiducial points
#input: index
#output: matrix of fiducial points corresponding to the index
readMat.matrix <- function(index){
     return(round(readMat(paste0(test_pt_dir, sprintf("%04d", index), ".mat"))[[1]],0))
}

#load fiducial points
fiducial_pt_list <- lapply(1:n_files, readMat.matrix)
save(fiducial_pt_list, file="../output/fiducial_pt_list_test.RData")

Step 3: construct features and responses

• The follow plots show how pairwise distance between fiducial points can work as feature for facial emotion recognition.

  • In the first column, 78 fiducials points of each emotion are marked in order.
  • In the second column distributions of vertical distance between right pupil(1) and right brow peak(21) are shown in histograms. For example, the distance of an angry face tends to be shorter than that of a surprised face.
  • The third column is the distributions of vertical distances between right mouth corner(50) and the midpoint of the upper lip(52). For example, the distance of an happy face tends to be shorter than that of a sad face.

Figure1

feature.R should be the wrapper for all your feature engineering functions and options. The function feature( ) should have options that correspond to different scenarios for your project and produces an R object that contains features and responses that are required by all the models you are going to evaluate later.

• feature.R
• Input: list of images or fiducial point
• Output: an RData file that contains extracted features and corresponding responses

source("../lib/feature.R")
#tm_feature_train <- NA
#if(run.feature.train){
#  tm_feature_train <- system.time(dat_train <- feature(fiducial_pt_list, train_idx))
#  save(dat_train, file="../output/feature_train.RData")
#}else{
#load(file="../output/feature_train.RData")
#}

tm_feature_test <- NA
if(run.feature.test){
  tm_feature_test <- system.time(dat_test <- feature(fiducial_pt_list, test_idx))
  save(dat_test, file="../output/feature_test.RData")
}else{
  load(file="../output/feature_test.RData")
}

Step 4: Train a classification model with training features and responses

Call the train model and test model from library.

train.R and test.R should be wrappers for all your model training steps and your classification/prediction steps.

• train.R
  • Input: a data frame containing features and labels and a parameter list.
  • Output:a trained model
• test.R
  • Input: the fitted classification model using training data and processed features from testing images
  • Input: an R object that contains a trained classifier.
  • Output: training model specification
• In this Starter Code, we use logistic regression with LASSO penalty to do classification.

#source("../lib/train_gbm.R")
source("../lib/test_gbm.R")
#source("../lib/fit_train_xgboost.R")

Run test on test images

*Baseline model

load("../output/res_cv_gbm.RData")
best_n.trees = as.numeric(res_cv_gbm[which.min(res_cv_gbm$mean_error),1])
best_shrinkage = as.numeric(res_cv_gbm[which.min(res_cv_gbm$mean_error),2])

source("../lib/test_gbm.R")
tm_test_baseline = NA
feature_test <- as.matrix(dat_test[, -6006])

if(run.test){
  load(file="../output/fit_train_baseline.RData")
  tm_test_baseline <- system.time({
    label_pred_baseline <- as.integer(test_gbm(fit_train_baseline,feature_test, best_n.trees, best_shrinkage, pred.type = 'link'));
    label_pred_baseline <- label_pred_baseline - 1;
  prob_pred_baseline <- test_gbm(fit_train_baseline, feature_test,best_n.trees, best_shrinkage, pred.type = 'response')})
}

*Prediction by baseline

predict_baseline <- as.data.frame(label_pred_baseline)
#write.csv(baseline, "label.csv", row.names = F)

*XGBoost (Improved model)

tm_test_xgb = NA
feature_test <- as.matrix(dat_test[, -6006])
colnames(feature_test) <- NULL
if(run.test){
  load(file="../output/fit_train_xgb.RData")
  tm_test_xgb <- system.time({label_pred_xgb <- predict(fit_train_xgb, feature_test, pred.type = 'class');
                          label_pred_xgb[label_pred_xgb >= 0.5] <- 1;
                          label_pred_xgb[label_pred_xgb < 0.5] <-  2;
                          label_pred_xgb  <- label_pred_xgb - 1
                          prob_pred_xgb <- predict(fit_train_xgb, feature_test, pred.type = 'response')})
}

*Prediction by Improved model

predict_xgb <- as.data.frame(label_pred_xgb)

*Combining the datasets into one and writing the .csv

index <- as.data.frame(1:2000)
output_csv <- cbind(index, predict_baseline, predict_xgb)
names(output_csv) <- c("index", "baseline model", "xgboost (improved) model")
write.csv(output_csv, "test_prediction.csv", row.names = F)

Evaluation

*Baseline Model

## reweight the test data to represent a balanced label distribution
label_test <- as.integer(dat_test$label)
weight_test <- rep(NA, length(label_test))
for (v in unique(label_test)){
  weight_test[label_test == v] = 0.5 * length(label_test) / length(label_test[label_test == v])
}

accu_baseline <- mean(label_pred_baseline == label_test)
tpr.fpr.baseline <- WeightedROC(prob_pred_baseline, label_test, weight_test)
auc_baseline <- WeightedAUC(tpr.fpr.baseline)


cat("The accuracy of model GBM: with n.trees=",best_n.trees,"and shrinkage =", best_shrinkage, "is", accu_baseline*100, "%.\n")
cat("The AUC of model GBM: with n.trees=", best_n.trees,"and shrinkage =", best_shrinkage, "is", auc_baseline, ".\n")

*Improved Model

label_test <- as.integer(dat_test$label)
label_test_xgb <- label_test
label_test_xgb[label_test_xgb==2] = 0

weight_test <- rep(NA, length(label_test_xgb))
for (v in unique(label_test_xgb)){
  weight_test[label_test_xgb == v] = 0.5 * length(label_test_xgb) / length(label_test_xgb[label_test_xgb == v])
}

accu_xgb <- mean((label_pred_xgb == label_test_xgb))
tpr.fpr_xgb <- WeightedROC(prob_pred_xgb, label_test_xgb, weight_test)
auc_xgb <- WeightedAUC(tpr.fpr_xgb)


cat("The accuracy of the XGBoost model:", "is", accu_xgb*100, "%.\n")
cat("The AUC of the XGBoost model:", "is", auc_xgb, ".\n")

Summarize Running Time

Prediction performance matters, so does the running times for constructing features and for training the model, especially when the computation resource is limited.

*Baseline Model

cat("Time for constructing training features=", tm_feature_train[1], "s \n")
cat("Time for constructing testing features=", tm_feature_test[1], "s \n")
cat("Time for training model=", tm_train_baseline[1], "s \n") 
cat("Time for testing model=", tm_test_baseline[1], "s \n")

• Improved Model

cat("Time for constructing training features=", tm_feature_train[1], "s \n")
cat("Time for constructing testing features=", tm_feature_test[1], "s \n")
cat("Time for training model=", xgb_train_time[1], "s \n") 
cat("Time for testing model=", tm_test_xgb[1], "s \n")

###Reference - Du, S., Tao, Y., & Martinez, A. M. (2014). Compound facial expressions of emotion. Proceedings of the National Academy of Sciences, 111(15), E1454-E1462.